The ABC of Computational Text Analysis
#1 Introduction +
Where is the digital revolution?
Alex Flückiger
Faculty of Humanities and Social Sciences
University of Lucerne
23 February 2023
Outline
- digital revolution or hype?
- about us
- goals of this course
AI: A non-standard Introduction
The World has changed, hasn’t it?
An Era of Big Data + AI
Group Discussion
What makes a computer looking intelligent?
AI is a moving target with respect to …
- human capabilities
- technological abilities
Transfer of Human Intelligence
from static machines to more flexible devices
- mimicking intelligent behavior
- reading + seeing + hearing
- speaking + writing + drawing
- a sense of contextual perception
- many degrees of freedom
Seeing like a Human?
An image segmentation by Facebook’s Detectron2 (Wu et al. 2019)
Speaking like a Human?
Speech-to-Text (STT)
Recognizing speech regardless of language, accent, speed, noise etc.
Check out samples of Whisper (Radford et al. 2022)
Check demo for Swiss German (Plüss et al. 2021)
Text-to-Speech Synthesis (TTS)
Personalizing voice given an audio sample of 3s
- Check out samples of VALL-E (Wang et al. 2023)
Generative and Multimodal AI
Outsmarting Humans?
ChatGPT is amazing but …
… it is also a stochastic parrot. 🦜
(Bender et al. 2021)
Can you disenchant ChatGPT?
Experiment with ChatGPT
- What works (surprisingly) well?
- When does it fail?
Trend towards Multimodality
Breakthrough by combining language processing and image generation with Muse (Chang et al. 2023)
Deepfakes? Yes, they are real!
Editing pictures with Muse using natural language (Chang et al. 2023)
Video is just the last barrier…
Synthesize any content with ever increasing quality
- Checkout this demo trailer for authentic dubbing
- Use words and images to synthesize new videos with Gen-1(Esser et al. 2023)
🎥
Artificial Intelligence
Subfields
- Natural Language Processing (NLP)
- Computer Vision (CV)
- Robotics
How does Computer Intelligence work?
- interchangeably (?) used concepts
- Artificial Intelligence (AI), Machine Learning (ML), Deep Learning (DL)
- learn patterns from lots of data
- more recycling than genuine intelligence
- theory agnostically
- supervised training is the most popular
- learn relation between input and output
AI is also Hype
AI is different to Human Intelligence
Why this matters for
Social Science
Computational Social Science
data-driven research
- computational social science (Lazer et al. 2009; Salganik 2017)
- Digital Humanities, Computational History, Data Science
- highly interdisciplinary
- machine learning empowers researchers (Lundberg, Brand, and Jeon 2022)
- early computational history already in 1960s (Graham, Milligan, and Weingart 2015)
Group Discussion
What kind of data is there?
What data is relevant for social science?
- data as traces of social behaviour
- tabular, text, image
- datafication
- sensors of smartphone, digital communication
- much of human knowledge compiled as text
About the Mystery of Coding
coding is like…
- cooking with recipes
- superpowers
Women have coding powers too!
Where the actual Revolution is
Coding is a superpower …
- flexible
- reusable
- reproducible
- inspectable
- collaborative
… to tackle complex problems on scale
About us
Personal Example
directed country mentions in UN speeches
Goals of this Course
What you learn
- collect and curate data
- computationally analyze, interpret, and visualize texts
- command line + Python
- digital literacy + scholarship
- problem-solving capacity
Learnings from previous Courses
- too much content, too little practice
- programming can be overwhelming
- learning by doing, doing by googling
Levels of Proficiency
- awareness of today’s computational potential
- analyzing existing datasets
- creating + analyzing new datasets
- applying advanced machine learning
How I teach
- computational practises
- critical perspective on technology
- lecture-style introductions
- hands-on coding sessions
- discussions + experiments in groups
Provisional Schedule
| Date | Topic |
|---|---|
| 23 February 2023 | Introduction + Where is the digital revolution? |
| 02 March 2023 | Text as Data |
| 09 March 2023 | Setting up your Development Environment |
| 16 March 2023 | Introduction to the Command-line |
| 23 March 2023 | Basic NLP with Command-line |
| 30 March 2023 (Zoom) | Learning Regular Expressions |
| 06 April 2023 (Zoom) | Working with (your own) Data |
| no lecture (Osterpause) | |
| 20 April 2023 | Ethics and the Evolution of NLP |
| 27 April 2023 | Introduction to Python + VS Code |
| 04 May 2023 | Data Analysis of Swiss Media |
| 11 May 2023 | NLP with Python |
| no lecture (Christi Himmelfahrt) | |
| 25 May 2023 | NLP with Python II + Working Session |
| 01 June 2023 | Mini-Project Presentations + Discussion |
🖥️ There are two digital lectures via Zoom.
TL;DR 🚀
You will be tech-savvy…
…yet no programmer applying fancy machine learning
Requirements
- no technical skills required ✅
- self-contained course
- laptop (macOS, Win11, Linux) 💻
- update system
- free up at least 15GB storage
- backup files
Grading ✍️
- 3 exercises during semester
- no grades (pass/fail)
- mini-project with presentation
- backup claims with numbers
- work in teams
- data of your interest
- optional: writing a seminar paper
- in cooperation with Prof. Sophie Mützel
Organization
- seminar on Thursday from 2.15pm - 4.00pm
- additionally, streaming via Zoom
- course website KED2023 with slides + information
- readings on OLAT
- communication on OLAT Forum
- forum for everything except personal
- subscribe to notifications
- direct: alex.flueckiger@doz.unilu.ch
Who are you?
Please fill out this questionnaire
📝
Questions?
Reading
Required
Lazer, David, Alex Pentland, Lada Adamic, Sinan Aral, Albert-László Barabási, Devon Brewer, Nicholas Christakis, Noshir Contractor, James Fowler, Myron Gutmann, Tony Jebara, Gary King, Michael Macy, Deb Roy, and Marshall Van Alstyne. 2009. “Computational Social Science.” Science 323(5915):721–23.
(via OLAT)
Optional
Graham, Shawn, Ian Milligan, and Scott Weingart. 2015. Exploring Big Historical Data: The Historian’s Macroscope. Open Draft Version. Under contract with Imperial College Press.
References
Bender, Emily M., Timnit Gebru, Angelina McMillan-Major, and Shmargaret Shmitchell. 2021. “On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜.” In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, 610–23. Virtual Event Canada: ACM. https://doi.org/10.1145/3442188.3445922.
Chang, Huiwen, Han Zhang, Jarred Barber, A. J. Maschinot, Jose Lezama, Lu Jiang, Ming-Hsuan Yang, et al. 2023. “Muse: Text-To-Image Generation via Masked Generative Transformers.” arXiv. https://doi.org/10.48550/arXiv.2301.00704.
Esser, Patrick, Johnathan Chiu, Parmida Atighehchian, Jonathan Granskog, and Anastasis Germanidis. 2023. “Structure and Content-Guided Video Synthesis with Diffusion Models.” arXiv. https://doi.org/10.48550/arXiv.2302.03011.
Graham, Shawn, Ian Milligan, and Scott Weingart. 2015. Exploring Big Historical Data: The Historian’s Macroscope. Open Draft Version. Under contract with Imperial College Press. http://themacroscope.org.
Lazer, David, Alex Pentland, Lada Adamic, Sinan Aral, Albert-László Barabási, Devon Brewer, Nicholas Christakis, et al. 2009. “Computational Social Science.” Science 323 (5915): 721–23. https://doi.org/10.1126/science.1167742.
Lundberg, Ian, Jennie E. Brand, and Nanum Jeon. 2022. “Researcher Reasoning Meets Computational Capacity: Machine Learning for Social Science.” Social Science Research 108 (November): 102807. https://doi.org/10.1016/j.ssresearch.2022.102807.
Plüss, Michel, Lukas Neukom, Christian Scheller, and Manfred Vogel. 2021. “Swiss Parliaments Corpus, an Automatically Aligned Swiss German Speech to Standard German Text Corpus.” arXiv. https://doi.org/10.48550/arXiv.2010.02810.
Radford, Alec, Jong Wook Kim, Tao Xu, Greg Brockman, Christine McLeavey, and Ilya Sutskever. 2022. “Robust Speech Recognition via Large-Scale Weak Supervision.” arXiv. https://doi.org/10.48550/arXiv.2212.04356.
Salganik, Matthew J. 2017. Bit by Bit: Social Research in the Digital Age. Illustrated edition. Princeton: Princeton University Press. https://www.bitbybitbook.com.
Wang, Chengyi, Sanyuan Chen, Yu Wu, Ziqiang Zhang, Long Zhou, Shujie Liu, Zhuo Chen, et al. 2023. “Neural Codec Language Models Are Zero-Shot Text to Speech Synthesizers.” arXiv. https://doi.org/10.48550/arXiv.2301.02111.
Wu, Yuxin, Alexander Kirillov, Francisco Massa, Wan-Yen Lo, and Ross Girshick. 2019. Detectron2. Meta Research. https://github.com/facebookresearch/detectron2.